ARTICLE

DOI: 10.1038/s41467-017-01173-4 OPEN YEATS2 links to tumorigenesis of non-small cell lung cancer

Wenyi Mi 1,2, Haipeng Guan3,4, Jie Lyu 5, Dan Zhao3,4, Yuanxin Xi5, Shiming Jiang1,2, Forest H. Andrews6, Xiaolu Wang1,2, Mihai Gagea 7, Hong Wen1,2, Laszlo Tora 8,9,10,11, Sharon Y.R. Dent1,2,12, Tatiana G. Kutateladze6, Wei Li 5, Haitao Li 3,4 & Xiaobing Shi 1,2,12

Recognition of modified by “reader” constitutes a key mechanism reg- ulating diverse -associated processes important for normal and neoplastic devel- opment. We recently identified the YEATS domain as a novel acetyllysine-binding module; however, the functional importance of YEATS domain-containing proteins in human cancer remains largely unknown. Here, we show that the YEATS2 is highly amplified in human non-small cell lung cancer (NSCLC) and is required for cancer cell growth and survival. YEATS2 binds to acetylated via its YEATS domain. The YEATS2-containing ATAC complex co-localizes with H3K27 acetylation (H3K27ac) on the promoters of actively tran- scribed . Depletion of YEATS2 or disruption of the interaction between its YEATS domain and acetylated histones reduces the ATAC complex-dependent promoter H3K9ac levels and deactivates the expression of essential genes. Taken together, our study identifies YEATS2 as a histone H3K27ac reader that regulates a transcriptional program essential for NSCLC tumorigenesis.

1 Department of and Molecular Carcinogenesis, The University of Texas M.D. Anderson Cancer Center, Houston, TX 77030, USA. 2 Center for Cancer Epigenetics, The University of Texas M.D. Anderson Cancer Center, Houston, TX 77030, USA. 3 MOE Key Laboratory of Sciences, Beijing Advanced Innovation Center for Structural Biology, Department of Basic Medical Sciences, School of Medicine, Tsinghua University, Beijing 100084, China. 4 Tsinghua-Peking Joint Center for Life Sciences, Tsinghua University, Beijing 100084, China. 5 Department of Molecular and Cellular Biology, Dan L. Duncan Cancer Center, Baylor College of Medicine, Houston, TX 77030, USA. 6 Department of Pharmacology, University of Colorado School of Medicine, Aurora, CO 80045, USA. 7 Department of Veterinary Medicine & Surgery, The University of Texas M.D. Anderson Cancer Center, Houston, TX 77030, USA. 8 Institut de Génétique et de Biologie Moléculaire et Cellulaire, 67404 Illkirch, France. 9 Centre National de la Recherche Scientifique, UMR7104, 67404 Illkirch, France. 10 Institut National de la Santé et de la Recherche Médicale, U964, 67404 Illkirch, France. 11 Université de Strasbourg, 67404 Illkirch, France. 12 Genes and Development and Epigenetics & Molecular Carcinogenesis Graduate Programs, The University of Texas Graduate School of Biomedical Sciences, Houston, TX 77030, USA. Wenyi Mi, Haipeng Guan, Jie Lyu and Dan Zhao contributed equally to this work. Correspondence and requests for materials should be addressed to W.L. (email: [email protected]) or to H.L. (email: [email protected]) or to X.S. (email: [email protected])

NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4

ysine acetylation is one of the most frequent post- ovarian serous cystadenocarcinoma (27%), and head and Ltranslational modifications occurring on histones that play neck squamous cell carcinoma (23%) (Fig. 1a). Importantly, a critical role in regulating chromatin dynamics and the YEATS2 gene expression levels are positively correlated to its accessibility of the underlying DNA in eukaryotes1. Acetylation amplification status in these tumors (Supplementary Fig. 1b–d). on histone residues is controlled by two families of coun- In human NSCLC and ovarian cancer patients, high YEATS2 teracting : histone acetyltransferases (HATs) and histone expression levels are correlated with worse prognosis (Supple- deacetylases (HDACs), and is normally associated with active mentary Fig. 1e, f). transcription2, 3. In addition to neutralizing the positive charge on We next assessed YEATS2 expression levels across a number of the side chain of lysine residues, the bulky acetyl groups can also lung cancer cell lines. Compared to the immortalized “normal” serve as docking sites for reader proteins, which recognize this lung fibroblast cell lines (WI-38 and IMR-90), YEATS2 was specific modification and transduce the molecular signals to elicit overexpressed at both transcript and protein levels in all NSCLC various downstream biological outcomes4. Bromodomain (BRD) cell lines we examined (Fig. 1b and Supplementary Fig. 2a). has long been thought to be the sole protein module that speci- YEATS2 is a stoichiometric component of the ATAC HAT fically recognizes acetyllysine motifs5. Some tandem plant complex, which catalyzes histone acetylation, mainly on H3K9 homeodomain zinc fingers were later found to bind histone H3 in and H3K14, by the enzymatic subunit GCN5 or PACF12, 13. – an acetylation-sensitive manner6 8. Recently, we identified the Interestingly, compared with the immortalized normal cells, we YEATS domain of AF9 protein as a novel reader of histone also observed elevated levels of GCN5 and PCAF in most acetylation9. YEATS domain is evolutionarily conserved from examined lung cancer cells (Fig. 1b), suggesting that essential yeast to human10. There are four YEATS domain-containing subunits of the ATAC complex cooperate in human cancers likely proteins in humans and three in Saccharomyces cerevisiae11. All leading to an super-active complex. Consistent with this the YEATS domain proteins are associated with chromatin- speculation, we found global histone acetylation levels, especially associated complexes, such as HAT complexes and chromatin- H3K9ac, were evidently higher in the NSCLC cell lines than the remodeling complexes, however, the functions of these proteins, immortalized normal cells (Fig. 1b). Interestingly, we also and particularly their YEATS domains, are not well understood. observed increased HDAC1 protein levels in cancer cells, which YEATS domain-containing 2 (YEATS2) is a scaffolding sub- is opposite to the increased K9 acetylation (Fig. 1b). unit of the Ada-two-A-containing (ATAC) complex, a conserved Even though cancer cells acquire multiple genetic and metazoan HAT complex12, 13. Vertebrate ATAC complexes share epigenetic abnormalities, their growth and survival are often the same catalytic HAT subunit, GCN5, or the highly related impaired by inactivation of a single oncogene. Since YEATS2 is PCAF in mammals, with another multi-subunit complex highly amplified in NSCLC, we sought to determine whether Spt–Ada–Gcn5–acetyltransferase (SAGA)14, 15. Although the depletion of YEATS2 affects lung cancer cell growth. To this end, SAGA complex has been extensively studied in both yeast and we knocked down (KD) YEATS2 gene expression in the H1299 humans, much less is known about the ATAC complex. GCN5 lung adenocarcinoma cell line using two independent shRNAs and PCAF in the ATAC complex mainly acetylate histone H3K9 (Supplementary Fig. 2b) and determined cell growth. We and H3K14, while the second acetyltransferase ATAC2 in the observed a marked suppression of cell proliferation in cells complex has been reported to modify H4K1616, 17. The ATAC treated with YEATS2-targeting shRNAs (shY2) compared with complex occupies distinct set of genes from SAGA and coordi- the cells treated with a non-targeting control shRNA (shNT) nates MAP kinases to regulate JNK target genes18, 19. The sub- (Fig. 1c). Notably, the levels of suppression were correlated with units of SAGA form four sub-modules that exert distinct the KD efficiency, with severe growth defect observed in the cells molecular functions within the complex20, 21, however, within the with higher KD efficiency. The growth inhibition by YEATS2 KD ATAC complex, except the HAT module, the functions of most was also observed in additional NSCLC cell lines (A549, H520, other subunits remain largely unknown. In this study, we char- and Ludlu-1) and ovarian cancer cell lines (CaoV3 and HeyA8) acterized the molecular and biological functions of YEATS2 that also harbor YEATS2 amplification, as well as in the within the ATAC complex. We found that the YEATS2 gene is immortalized normal lung fibroblast cells (WI-38 and IMR-90) highly amplified in human cancers including non-small cell lung that do not have YEATS2 overexpression (Supplementary cancer (NSCLC). Depletion of YEATS2-reduced cancer cell Fig. 2c–i), suggesting that YEATS2 is an essential gene for a growth, survival and transformation activity. The YEATS domain broad range of cancer cell lines as well as non-cancerous cells. of YEATS2 binds to acetylated histone H3K27 (H3K27ac). Cancer cells evolve with capability to undergo unlimited cell Recognition of histone acetylation is important for the functions division and transformation. We next sought to test whether of YEATS2 in cells. Disruption of acetylation recognition of YEATS2 is required for cell survival and transformation of YEATS2-abrogated GCN5/PCAF-mediated promoter histone NSCLC. In clonogenic assay of both H1299 and A549 cells, the acetylation and consequently, suppressed the expression of its YEATS2 KD cells developed fewer colonies compared with the target genes, including the ribosomal protein-encoding genes that control cells, suggesting that YEATS2 is required for lung cancer are essential for cell growth and survival. Taken together, our cell survival (Fig. 1d and Supplementary Fig. 2j). We also results identified YEATS2 as a histone H3K27ac reader that performed soft agar colony formation assays to determine the epigenetically regulates a transcriptional program essential for effect of YEATS2 KD on anchorage-independent growth, an NSCLC tumorigenesis. ability of transformed cells to grow independently of a solid surface22. Compared with the shNT treated control cells, YEATS2 KD resulted in fewer and also smaller colonies in soft Results agar in both H1299 and A549 cells (Fig. 1e and Supplementary YEATS2 is an essential gene amplified in NSCLC. To determine Fig. 2k). Importantly, the defects associated with YEATS2 KD whether YEATS2 plays a role in human cancers, we first exam- in both clonogenic and anchorage-independent cell growths ined YEATS2 gene expression status across cancers in The Cancer were rescued by ectopic expression of shRNA-resistant Genome Atlas database via The cBioPortal for Cancer Genomics. YEATS2. Taken together, these results indicate that YEATS2 is As part of the 3q26 amplicon (Supplementary Fig. 1a), YEATS2 is required for cell growth, survival, and transformation of lung highly amplified in a variety of human cancers, including cancer cells. lung squamous cell carcinoma (56% amplification frequency),

2 NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4 ARTICLE

YEATS2 controls the expression of ribosome protein genes.To Encyclopedia of Genes and Genomes pathway analysis of the determine how YEATS2 regulates cancer cell growth and survival, differentially expressed genes using DAVID (Database for we performed RNA-seq analysis in YEATS2 KD cells to identify Annotation, Visualization, and Integrated Discovery) revealed the genes regulated by YEATS2 genome-wide. We used YEATS2- that the dysregulated genes were involved in vital biological targeting shRNA shY2-1 since this shRNA exhibited an efficient processes, with downregulated genes enriched in the pathways KD (Fig. 1c and Supplementary Fig. 2b), and we performed RNA- regulating ribosome biogenesis, DNA replication, cell cycle, DNA seq experiments in duplicates. We identified 1748 genes that were repair, and splicing, whereas upregulated genes enriched in the downregulated (false discovery rate (FDR) < 0.01), whereas 3361 pathways of lysosome functions, glycan degradation, and focal genes upregulated, in YEATS2 KD cells compared with the adhesion, etc (Fig. 2b and Supplementary Data 3). RNA-seq control cells (Fig. 2a and Supplementary Data 1, 2). Kyoto analysis using an independent shRNA (shY2-2) that partially KD

abLung Lung cancer H3K9ac/H3 normal 3 ** cells cells * 60% Amplification 2 ** *

Mutation N. S. * Folds 1 Deletion 50% WI38 IMR90 A549 H1299 H1355 Ludlu-1 H520 0 Multiple alterations YEATS2 150 KDa 40% H3K14ac/H3 GCN5 100 KDa * N. S. 2 N. S. N. S. PCAF 100 KDa N. S. 30% * 1 H3K9ac 15 KDa Folds 20% H3K14ac 15 KDa 0 H3K27ac 15 KDa H3K27ac/H3 3 Alteration frequency (TCGA) 10% ** H3 15 KDa 2 ** ** *

75 KDa N. S. 0% HDAC1 *

Folds 1 Actin NCI-60 CCLE 37 KDa 0

Ovarian (TCGA) Cervical (TCGA) Uterine (TCGA) Lung squ (TCGA) EsophagusUterine (TCGA)Ovarian CS (TCGA)(TCGA pub)UterineLung (TCGA adeno pub) (TCGA) Lung squ (TCGAHead pub) & neck (TCGA) Head & neck (TCGA pub) c e H1299 shNT 40 shY2-1 shNT+VectorshY2+Vector shY2+YEATS2 shY2-2 ) 4 30 10 × * 20 *** *** ** 400 *** 200 *** N.S. m) N.S. μ

shNT shY2-1 10 shY2-2 Cell number ( 300 150 YEATS2 Actin 0 200 100 01234 Days 100 50 Colony diameter (

d Colony numbers per field 0 0 shNT+VectorshY2+Vector shY2+YEATS2

shNT+Vector shY2+Vector shY2+YEATS2 shNT+VectorshY2+Vector shNT+VectorshY2+Vector shY2+YEATS2 shY2+YEATS2 YEATS2 Actin

Fig. 1 YEATS2 is amplified in NSCLC and is required for cancer cell growth and survival. a YEATS2 gene is frequently amplified in various types of human cancers. Data was obtained from the cBioPortal for Cancer Genomics. b Western blot analysis of YEATS2, GCN5, PCAF, HDAC1, and the indicated histone acetylation in NSCLC cell lines and immortalized “normal” lung fibroblast cell lines. Total H3 and actin are shown as loading control. The arrow indicates the band of YEATS2 protein. Relative H3K9ac, H3K14ac, and H3K27ac levels were quantified (n = 3, mean ± s.e.m.). N.S. not significant; *p < 0.05; **p < 0.01 (Student’s t-test). c Cell proliferation assay of H1299 cells treated with control (shNT) or YEATS2 shRNAs (shY2). Cells (mean ± S.E.M., n = 4) were counted for 4 days after seeding (left panel). Right panel: western blot analysis showing YEATS2 knockdown efficiency. The arrow indicates the band of YEATS2 protein. Error bars represent S.E.M. *p < 0.05; **p < 0.01 (Student’s t-test). d Clonogenic assay of control (shNT), YEATS2 knockdown (shY2), and knockdown H1299 cells rescued with ectopic expression of YEATS2. Empty vector was used as a control. Colonies were stained and photographed 7 days after seeding (right panel). Left panel: western blot analysis of YEATS2 expression level in indicated cells. The arrow indicates the band of YEATS2 protein. e Anchorage-independent growth assay of H1299 cells as in (d). Cells (mean ± S.E.M., n = 4–6) were stained with 0.005% crystal violet blue and photographed 3 weeks after seeding (top panel). Colony numbers (bottom left) and colony diameters (bottom right) were measured and quantified using ImageJ software. Scale bar, 200 µm. N.S. not significant; ***p < 0.001 (Student’s t-test)

NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4

abshNT shY2 Ribosome (48)

DNA replication (29)

Cell cycle (40) (1748) 1748 genes 15 Mismatch repair (15)

Spliceosome (36) Downregulated genes Downregulated

10 Lysosome (57)

Other glycan degradation (10) 5 Glycosaminoglycan degradation (11) 3361 genes ECM-receptor interaction (3361) 0 (25)

Focal adhesion (47) Upregulated genes

01020

–log10P cd

RPL27 shNT RPS15 RPL7 shY2-1 20 RPL7A shY2-2 RPL38 RPL29 * * ** ** * *** ** ** 1.2 ** ** ** ** ** ** ** ** ** RPL35 1.0 0.8 RPL26 10 0.6

–log10(FDR) RPL6 0.4

RPL8 0.2 Relative mRNA level Relative 0.0

0 RPL6 RPL7 RPL8 –6 –3 0 3 6 RPL7A RPS15 RPL26 RPL27 RPL29 RPL35 RPL38 log2 fold change

Fig. 2 YEATS2 is required for the expression of ribosome protein-encoding genes. a Heatmap representation of differentially expressed genes in control (shNT) and YEATS2 knockdown (shY2) cells from two independent biological replicates of RNA-seq experiments. Fisher’s exact test was used to define differentially expressed genes (q < 0.01). The color key represents normalized Log2 expression values. b Kyoto Encyclopedia of Genes and Genomes pathway analysis of downregulated (green) or upregulated (red) genes in YEATS2 knockdown cells compared with control cells. The numbers of genes within each functional group are shown in parenthesis. Fisher’s exact test was used to identify the biological function with significant p-values (Benjamini–Hochberg corrected p < 0.05). c Volcano plot of differentially expressed genes in YEATS2 knockdown cells compared with control cells. 49 downregulated ribosomal protein genes are shown in green and 30 non-differentially expressed ribosomal protein genes in black. FDR, false discovery rate. d Quantitative RT-PCR (qRT-PCR) analysis of the expression of ten randomly picked ribosomal protein genes in control (shNT) and YEATS2 knockdown (shY2) cells. Error bars indicate S.E.M. of three biological replicates. *p < 0.05; **p < 0.01 (Student’s t-test)

YEATS2 identified 1620 genes downregulated, among which little or no defect in cell apoptosis or migration (Supplementary significant number of genes (520), including 11 ribosomal protein Fig. 3c–e), suggesting that growth suppression by YEATS2 KD is, genes, overlapped in both KD cells (Supplementary Fig. 3a, b). at least in part, due to perturbation of cell cycle progression and Notably, both downregulated and upregulated genes were enri- DNA replication. Taken together, these results suggest that ched in pathways in cancers, including lung, colorectal, and YEATS2 regulates the expression of genes involved in critical pancreatic cancers (Supplementary Data 3), suggesting that pathways such as ribosome biogenesis that are essential for YEATS2 controls growth and survival of various types of tumors maintaining cell growth and survival. through transcriptional regulation of essential genes. Because YEATS2 is a subunit of the ATAC complex that is mostly involved in gene activation, we first focused on the genes The YEATS domain of YEATS2 binds to acetylated H3K27. downregulated by YEATS2 depletion in this study. The ribosome The recognition of histone H3 acetylation is an evolutionarily is a cellular machine for protein synthesis that is essential for conserved function of the AF9 YEATS domain9, we thus reasoned sustained growth of both normal and cancer cells. Strikingly, that the YEATS domain of human YEATS2 (Fig. 3a) may also among the genes encoding all 79 known ribosomal proteins, 49 binds to acetylated histones. To test this hypothesis, we per- genes were downregulated whereas none were upregulated in cells formed histone peptide pulldown assays and we found that the treated with YEATS2 shRNA (shY2-1) (Fig. 2c and Supplemen- YEATS domain of YEATS2 bound specifically to histone tary Data 1 and 2). Downregulation of these ribosomal protein- H3K27ac, with weak or no bindings to other acetylated histone encoding genes was validated by qRT-PCR in cells treated with peptides (Fig. 3b). Histone-binding assay in vitro and protein- two YEATS2-targeting shRNAs, with levels of suppression chromatin binding in cells demonstrated the interaction between correlated with KD efficiency (Fig. 2d). Flow cytometry analyses YEATS2 and H3K27ac at full-length histone and nucleosomal revealed that KD of YEATS2 led to G1 arrest of cell cycle, whereas levels, respectively (Fig. 3c, d). Quantitative isothermal titration

4 NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4 ARTICLE

a 1231 310 1422 cdInput Flag IP

YEATS2 YEATS HF YEATS 10% input GST YEATS2 Vector Flag-YEATS2 Vector Flag-YEATS2 b GST Flag 10% input H3 (1–22) H3K4ac H3K9ac H3K14ac H3K18ac H3 (18-42) H3K23ac H3K27ac YEATS2YEATS H3K27ac H3K27ac

10% input H4 (1-23) H4K5ac H4K8ac H4K12ac H4K16ac 10% input H2B (1-22) H2BK5ac H2BK12ac H2BK15ac H2BK20ac f g L1 C K27ac A31 e K27ac A29 1.0 L8 P30 0.5 L4 L6

) 0.0 S28 β A25

–1 2 –0.5 N β5 –1.0 β4 β8 A24 R26 β1 YEATS2 β –1.5 β6 7 KD(mM) L5 β3 –2.0 H315–39K27ac 0.05 W282 –2.5 H31–25K18ac 0.12 L2’ L3’ 3.3 Injection (kcal mol H31–34un 4.15 3.5 –3.0 α1’ H36–15K14ac N.D. L7 –3.5 H31–15K9ac N.D. 3.5 kT/e –4.0

0.0 0.5 1.0 1.5 2.0 2.5 C Y262 –10 10 N Molar ratio

h –1.2 1.8

A31 K27ac j Y313 A29 k P30 S28 W282 G281 A25 A31 Y2YEATS R26 10% input H3 (18–42) H3K27ac K27ac A29 A24 WT S261 G283 P30 H259A L309 S261A S28 Y262 R311 Y262A i W282 H259 W234 E284 G281S G281 S230 P260 A25 F285 W282A S261 F285A R26 H259 K27ac A24 Y262 F285

Fig. 3 The YEATS domain of YEATS2 recognizes H3K27ac. a Schematic representation of YEATS2 . The amino acid numbers of the YEATS domains and the full-length protein are shown. HF: histone fold domain. b Western blot analysis of histone peptide pulldowns of GST-YEATS2 YEATS domain and the indicated biotinylated peptides. c Western blot analysis of histone pulldowns of GST-YEATS2 YEATS domain or GST and calf thymus histones. The arrow indicates the band of the GST-YEATS YEATS2 domain. d Western blot analysis of Flag co-IP in 293 T cells transfected with

Flag-YEATS2 or vector control. e ITC fitting curves of YEATS2 YEATS titrated with H315–39K27ac, H31–15K18ac, unmodified H31–34K27, H36–15K14ac, and H31–15K9ac peptides. f Overall structure of YEATS2 (aa201–332) bound to the H324–31K27ac peptide in ribbon view. YEATS2 (pale cyan) is shown as ribbons, and the histone H3 peptide (yellow) is depicted as sticks. Purple mesh: Fo–Fc omit map around H324–31K27ac peptide contoured at 1.8 σ level. Bottom right, close-up view of the Kac-sandwiching pocket; interplanar distances are labeled in the unit of angstrom. g YEATS2–H3K27ac space-filling- surface view color-coded by electrostatic potential ranging from −10 to 10 kT/e. h Conservation mapping around the H3-binding surface in YEATS2. White and cyan colors indicate low (<0.25) and high (1.0) sequence conservation, respectively. The H3K27ac peptide is shown in yellow stick. i Close-up view of the K27ac-binding pocket of the YEATS2 YEATS domain. The pocket is displayed as semi-transparent surface with key residues shown as green sticks. K27ac is depicted in both yellow stick and space-filling sphere modes. j Hydrogen bonding network between H3K27ac peptide and YEATS2. Hydrogen bonds are shown as pink dashes. Key residues of YEATS2 are depicted as green sticks and labeled black; the H3 peptide is shown as yellow sticks and labeled red. k Western blot analysis of peptide pulldowns of WT YEATS2 YEATS domain or the indicated point mutants with the H3K27ac peptide

NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4

“ Table 1 Data collection and refinement statistics proper registration of H3K27ac. A recognition signature of Kac- X-X-Pro” is unique to H3K27ac but not H3K9ac despite that both sites share a consensus “A-R-Kac-S” motif, which explains the YEATS –H3 – K27ac YEATS2 24 31 binding specificity of YEATS2 to H3K27ac. The selectivity of the Data collection YEATS domain of YEATS2 was further validated by NMR Space group I422 experiments. 1H, 15N heteronuclear single-quantum coherence Cell dimensions 15 a, b, c (Å) 72.8, 72.8, 125.2 (HSQC) spectra of the uniformly N-labeled YEATS domain α, β, γ (°) 90, 90, 90 showed global chemical shift perturbations upon gradual addition Wavelength (Å) 0.9791 of the H3K27ac peptide, while H3K9ac peptide failed to induce Resolution (Å) 50–2.7 (2.79–2.70)* significant chemical shift changes in the protein (Supplementary Rmerge (%) 13.3 (85.9) Fig. 4d). Together, these results demonstrate YEATS2 is a histone I/σI 17.53 (2.71) H3K27 acetylation reader. Completeness (%) 99.5 (95.9) The HSQC results indicate that the residues surrounding Redundancy 3.8 (3.6) K27ac in the H3 peptide likely also contribute to the interaction Refinement (F > 0) between H3 and the YEATS2 YEATS domain. Analysis of the Resolution (Å) 39.8–2.7 peptide–protein interaction using LIGPLOT program also fl No. of re ections (test set) 4915 (467) revealed that the H324–31K27ac peptide is stabilized by a R R work/ free (%) 23.0/26.3 hydrogen bonding network and hydrophobic interactions invol- No. of atoms ving a number of residues including H259, S261, Y262, W282, Protein 1113 G283, E284, F285, and Y313 (Supplementary Fig. 4e). Indeed, Ligand 56 alanine mutation of the sandwich pocket residues completely Water 13 B-factors (Å2) disrupted the binding, highlighting their essential role for Protein 51.7 H3K27ac recognition (Fig. 3k; Supplementary Fig. 4f and Ligand 41.7 Supplementary Table 1). Besides, a 2-fold drop of histone P30A Water 44.4 mutation demonstrated the requirement of flanking amino acids R.m.s. deviations of H3K27 in mediating the YEATS2–H3K27ac interaction. Bond lengths (Å) 0.005 Bond angles (°) 1.08

*Values in parentheses are for the highest-resolution shell The ATAC complex co-localizes with H3K27ac and H3K9ac. The in vitro binding and structural data prompted us to deter- mine whether YEATS2 co-localizes with acetylated histone calorimetry (ITC) analysis revealed dissociation constant (KD)of H3K27 in cells. Extensive attempts to determine YEATS2 geno- 0.05 mM for the YEATS domain to the H3K27ac peptide, mic distribution using commercial YEATS2 antibodies and tag- 0.12 mM to the H3K18ac peptide, and weak or no binding to the ged ectopic YEATS2 failed. Since YEATS2 is a stoichiometric H3K9ac, H3K14ac, or unmodified histone peptides (Fig. 3e and component of the ATAC complex12, 13, 15, we performed chro- Supplementary Table 1). matin immunoprecipitation (ChIP) experiments followed by To decipher the underlying molecular basis for recognition of high-throughput sequencing (ChIP-seq) using a validated ChIP- histone acetylation by the YEATS domain of YEATS2, we seq grade antibody raised against another ATAC-specific subunit, crystallized the YEATS domain (aa 201–332) bound to the ZZZ318 to represent the genome-wide distribution of the ATAC H324–31K27ac peptide and solved the co-crystal structure at 2.7 Å complex. (Table 1). The overall structure of the YEATS domain adopts an High-throughput sequencing of ZZZ3 ChIP experiments immunoglobin β-sandwich fold between eight antiparallel β performed in duplicate identified 949 confident ZZZ3-bound strands (Fig. 3f). We modeled all 132 residues of YEATS2 YEATS peaks in H1299 cells (Supplementary Data 4). Notably, the domain and traced the “A24-A25-R26-K27ac-S28-A29-P30-A31” majority of ZZZ3 peaks resided within promoter regions (82.0%), residues of the H324–31K27ac peptide according to the electron and only small fractions localized in the transcribed regions density map. YEATS2 uses an aromatic sandwich cage for Kac (7.9%) or intergenic regions (10.1%) likely enhancers (Fig. 4a). As recognition with the acetylamide group of Kac clamped by YEATS2 specifically recognizes H3K27ac and the ATAC complex aromatic residues Y262 and W282 (Fig. 3g). The histone peptide- modifies H3K9ac, we also performed H3K27ac and H3K9ac binding surface of YEATS2 formed by loops L3, L5, and L7 ChIP-seq, which revealed 26,594 and 22,395 peaks, respectively (corresponding to loops L4, L6, and L8 of AF9) is less negative (Supplementary Data 5, 6). ZZZ3 was highly co-localized with compared to that of the AF9 YEATS domain9, which may partly acetylated histone H3; more than 90% of the ZZZ3-binding sites account for the relatively weak histone-binding activity observed were also co-occupied by both H3K27ac and H3K9ac (Fig. 4b). for YEATS2. The heatmap and average distribution of all ZZZ3 ChIP-seq Residue conservation analysis among YEATS2 YEATS ortho- peaks across transcription units revealed a strong enrichment at logs in various species reveals strict conservation of the crucial regions ±1 kb of transcription start sites, largely overlapping with amino acids that compose the H3K27ac-binding pocket (Fig. 3h, i the genomic distribution of H3K27ac and H3K9ac (Fig. 4c, d). and Supplementary Fig. 4a). In the complex structure of Genome browser views of the ChIP-seq signals of individual YEATS2–K27ac, the H3 peptide is stapled into the YEATS ZZZ3-bound genes and ChIP experiments followed by quantita- domain in an opposite N-to-C orientation compared to that of tive real-time PCR (ChIP-qPCR) analysis further confirmed the AF9 (Supplementary Fig. 4b, c). In the complex structure of AF9- co-localization of ZZZ3 with H3K27ac and H3K9ac in gene K9ac, the N-terminal motif “K4-Q5-T6-A7-R8” of H3 contributes promoters (Fig. 4e, f). to binding whereas in the complex structure of YEATS2–K27ac, ZZZ3 occupies enhancer regions in GM12878 lymphoblast the C-terminal motif “S28-A29-P30-A31” of H3 participates in cells and HeLa cells18. To determine whether ZZZ3 also binds to YEATS2 recognition. Notably, H3P30 at +3 position is anchored enhancers in lung cancer cells, we further performed ChIP-seq to at a hydrophobic pocket of YEATS2 (Fig. 3j), which promotes assess the genome-wide distribution of chromatin marks known to be associated with active promoters (H3K4me3) or with

6 NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4 ARTICLE

ab ZZZ3 ChIP-seq peaks H3K27ac

Genebody Intergenic 26 75 (7.9%) 96 (10.1%) 5291 ZZZ3 33 858 32 14,752

Promoter 778 (82.0%) 3025 H3K9ac c ZZZ3 H3K27ac H3K9ac H3K4me3

d 7 ZZZ3 H3K27ac 40 6 H3K9ac 40 5 30 30

20 4

10 20 3 0 ChIP-seq occupancy 10 2 Nomalized H3K9ac and H3K27ac

1 Normalized ZZZ3 ChIP-seq occupancy

ZZZ3 peaks enriched with active promoter marks –5 K TSS 20% 40% 60% 80% TES 5 K Others –5 K 5 K

e 20 5 K 5 K 5 K 5 K 5 K ZZZ3 100 H3K27ac 150 H3K9ac 0 RPL8 RPL27 RPL29 RPL35 RPL38

f 5.5 RPL6 RPL7 RPL7A RPL8 3.0 RPS15 RPL26 RPL27 0.5 RPL29 RPL35 0.04 RPL38 ChIP (% input) 0.02

0.00 IgG ZZZ3 H3K27ac H3K9ac

Fig. 4 The ATAC complex co-localizes with promoter H3K27ac genome-wide. a Genomic distribution of ZZZ3 ChIP-seq peaks in H1299 cells. The peaks are enriched in the promoter regions (transcription start site ±3kb). p < 2.2 × 10−16 (binomial test). b Venn diagram showing the overlap of ZZZ3 (blue), H3K27ac (red) and H3K9ac (green) occupied peaks. p < 1.79 × 10−63 (Super exact test). c Heatmaps of normalized density of ZZZ3, H3K27ac, H3K9ac and H3K4me3 ChIP-seq tags centered on ZZZ3-binding peaks in a ±5 kb window. The color key represents the signal density, where darker red represents higher ChIP-Seq signal. d Average genome-wide occupancies of ZZZ3 (blue), H3K27ac (red) and H3K9ac (green) along the transcription unit. The gene body length is normalized by percentage from the TSS to transcription termination site (TES). 5 kb regions upstream of TSS and 5 kb regions downstream of TES are also included. e Genome-browser view of the ZZZ3-ChIP-seq (blue), H3K27ac-ChIP-seq (red), and H3K9acChIP-seq (green) peaks on the indicated ribosomal protein genes. f qPCR analysis of ZZZ3, H3K27ac and H3K9ac ChIP in the promoters of representative ribosomal protein genes. IgG was used as a negative control. Error bars indicate S.E.M. of three biological repeats

NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4 enhancers (H3K4me1) in H1299 cells. The majority of the ZZZ3- indicate that YEATS2 is required for the recruitment of ATAC binding sites were enriched for H3K4me3, H3K9ac and H3K27ac complex to promoters and for ATAC-dependent maintenance of (Fig. 4c), whereas very few ZZZ3-binding sites overlapped with histone H3K9 acetylation on the ribosomal protein-encoding non-promoter H3K4me1, a mark of enhancers (Supplementary genes. Fig. 5a). Taken together, these results indicate that the ATAC complex co-localizes with acetylated histone H3 mainly on active The YEATS2 YEATS domain is required for tumor cell sur- promoters in H1299 cells. vival. Next, we asked whether the recognition of H3 acetylation by the YEATS domain is required for the function of YEATS2 in chromatin and transcriptional regulation. To address this ques- YEATS2 is required for ATAC-dependent maintenance of tion, we performed “rescue” experiments by ectopically expres- H3K9ac. GCN5 and PCAF in the ATAC and SAGA complexes sing shRNA-resistant WT YEATS2 or the acetylation-binding acetylate histone mainly on H3K9 and H3K14 to promote gene deficient mutants (Y262A and W282A) in YEATS2-depleted activation15, 16, 23. However, it still remains unknown how the H1299 cells (Supplementary Fig. 6a). We first assessed H3K9ac ATAC complex is recruited to specific chromatin loci that are and the expression levels of target genes, and we found that distinct from SAGA-bound regions. Because YEATS2 is an depletion of endogenous YEATS2 reduced H3K9ac and the ATAC-specific subunit and binds to H3K27ac, we hypothesized expression level of ribosomal protein-encoding genes (Fig. 6a, b). that YEATS2 recruits the ATAC complex to H3K27ac-enriched Importantly, ectopic expression of WT YEATS2, but not the target genes to promote active transcription via maintaining Y262A and W282A mutants, in the YEATS2-depleted cells promoter histone H3K9/H3K14 acetylation levels. If this restored H3K9ac on target gene promoters to levels comparable hypothesis is correct, depletion of YEATS2 should lead to dis- to those in the control cells (Fig. 6a). Consistently, WT YEATS2, sociation of the ATAC complex from chromatin and reduced but not the H3 acetylation-binding deficient mutants, rescued histone H3K9 and/or H3K14 acetylation levels. Indeed, immu- target gene expression in YEATS2 KD cells (Fig. 6b). noblot analysis of total histones in YEATS2 KD H1299 cells and We then sought to determine whether the YEATS domain is A549 cells revealed a marked reduction in H3K9ac levels, whereas required for YEATS2 function in regulating cell growth and only minor or no change in H3K14ac or H4K16ac levels (Fig. 5a), survival. We performed cell proliferation and colony formation suggesting that YEATS2 is required for maintaining global assays using the reconstitution system in which ectopic WT or H3K9ac levels. Interestingly, although GCN5 and PCAF are not mutant YEATS2 was reintroduced to the YEATS2 KD cells. reported as dominate HAT enzymes for H3K27ac, we also Consistent with the target gene expression patterns, ectopic observed marked reduction of H3K27ac levels upon YEATS2 KD expression of WT YEATS2 restored cell proliferation and colony (Fig. 5a). Nevertheless, stability of the ATAC complex compo- forming capability of the YEATS2-depleted cells, whereas the nents was not affected by YEATS2 KD, neither the HDAC1 Y262A and W282A mutants did not (Fig. 6c, d). Furthermore, in (Supplementary Fig. 5b). Next we asked whether YEATS2 is in vitro soft agar colony formation assays and in vivo xenograft required for ATAC-dependent histone H3K9ac on target gene assays, depletion of YEATS2-suppressed tumor growth. Impor- promoters. To this end, we performed H3K9ac ChIP-seq in both tantly, WT YEATS2, but not the H3 acetylation-binding deficient control and YEATS2 KD cells. Averaged H3K9ac ChIP-seq sig- mutants, restored the transformation capability of the YEATS2 nals revealed a moderate reduction of H3K9ac on the promoters KD H1299 cells in vitro and tumor growth in mice (Fig. 6e, f and of ZZZ3-occupied genes, whereas H3K9ac levels on non-ZZZ3- Supplementary Fig. 6b). Taken together, our results suggest a bound genes (others) remained largely unaffected (Fig. 5b). model wherein YEATS2 recognizes histone H3 acetylation and Consistent with the Western blot results (Fig. 5a), we also recruits the ATAC complex to chromatin, which in turn observed modest reduction of H3K27ac levels on ZZZ3-occupied maintains an open, acetylated chromatin environment to genes (Supplementary Fig. 5c). promote expression of genes essential for cancer cell proliferation, Comparison of the dysregulated genes by YEATS2 KD and the survival and tumorigenesis (Fig. 6g). ZZZ3-occupied genes suggested that only small number genes were direct targets of the YEATS2/ATAC complex (Supplemen- tary Fig. 5d), among which with 39 downregulated direct target Discussion genes enriched in the pathway of ribosome and in total only 10 Previously we identified the AF9/ENL YEATS domain as a his- upregulated genes enriched in two pathways (Supplementary tone acetylation reader module9. In addition to AF9/ENL, the two Fig. 5e, f and Supplementary Data 7). Genome browser views of functional paralogs that associate with the super elongation H3K9ac ChIP-seq signals and ChIP-qPCR analysis in control and complex or the DOT1L complex24, 25, humans have two other YEATS2 KD cells demonstrated reduction of H3K9ac levels on YEATS domain proteins, YEATS2 and YEATS4/GAS41, which the downregulated ribosomal protein-encoding genes in YEATS2 are components of the ATAC HAT complex and TIP60/SRCAP KD cells (Fig. 5c, d), whereas little or no changes in H3K9ac levels chromatin-remodeling complexes, respectively12, 16, 26, 27. Our on the 10 upregulated vesicular transport or lysosome genes biochemical and structural studies reveal that, different from the (Supplementary Fig. 5g). To determine whether the reduction of AF9 YEATS domain that recognizes acetylation on H3K9, K18 H3K9ac levels on the ribosomal proteins genes upon YEATS2 and K27, the YEATS domain of YEATS2 shows certain specifi- depletion is due to the dissociation of the ATAC complex from city, with H3K27ac as the best binding substrate whereas no chromatin, we performed ZZZ3 ChIP-seq and ChIP-qPCR detectable binding was observed to H3K9ac or H3K14ac. analyses in YEATS2 KD cells. Concurrent with changes in Nevertheless, the YEATS domain of YEATS2 utilizes aromatic H3K9ac levels, we observed reduced ZZZ3 occupancy on residues conserved among all YEATS domains forming a Ser/ individual ribosomal protein-encoding genes (Fig. 5c, e) as well Thr-lined sandwiching cage for encapsulation of the acetyl moi- as the averaged ChIP-seq signals of all ZZZ3-bound genes in ety. Together with our unpublished data of GAS41, our results YEATS2 KD cells (Fig. 5f). Again, in contrast, no significant demonstrate that recognition of histone acetylation is a common changes in ZZZ3 occupancy were observed on the upregulated feature of YEATS domains involving distinct chromatin- genes in YEATS2 KD cells (Supplementary Fig. 5h), indicating remodeling or histone-modifying complexes. that the upregulated genes in YEATS2 KD cells are likely not The hydrophobic pockets of all known YEATS domains are direct targets of the ATAC complex. Taken together, these results “open-ended”, enabling recognition of other types of acylation

8 NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4 ARTICLE

ab H1299 A549 150 ZZZ3 bound/shNT ZZZ3 bound/shY2 Others/shNT

shNT shY2

shNT shY2 Others/shY2 YEATS2 100 H3K9ac H3K14ac H3K27ac H3 50 H4K16ac H4–tetra–ac H4

Normalized H3K9ac occupancy 0

c –5 K –4 K–3 K–2 K–1 K TSS 1 K 2 K 3 K 4 K 5 K 5 K 5 K 5 K 5 K 5 K 400

H3K9ac-shNT 400 H3K9ac-shY2 0 20 ZZZ3-shNT 20 ZZZ3-shY2 0 RPL8 RPL27 RPL29 RPL35 RPL38

d ** 16 shNT shY2 ** 12 ** * * ******* f 8 8 ZZZ3 bound/shNT 4 ZZZ3 bound/shY2 H3K9ac ChIP 7 (Relative to H3) (Relative 0 Others/shNT 6 Others/shY2 RPL6 RPL7 RPL8 RPL7A RPS15 RPL26 RPL27 RPL29 RPL35 RPL38 5

e 4

3 0.06 * 2 ** ** ** ** * ** * ** * 0.04 Normalized ZZZ3 occupancy 1

0.02 –5 K –4 K –3 K –2 K –1 K TSS 1 K 2 K 3 K 4 K 5 K

0.00 ZZZ3 ChIP (% input)

RPL6 RPL7 RPL8 RPL7A RPS15 RPL26 RPL27 RPL29 RPL35 RPL38

Fig. 5 YEATS2 is required for ATAC-dependent maintenance of histone H3K9ac on the ribosomal protein genes. a Western blot analysis of YEATS2 and H3 and H4 acetylation in control (shNT) and YEATS2 KD (shY2) cells. H3 and H4 were used as a loading control. The arrow indicates the band of YEATS2 protein. b Average genome-wide H3K9ac occupancy on the promoter (5 kb±TSS) of the ZZZ3-bound genes or non-ZZZ3-bound genes (others) in control (shNT) and YEATS2 KD (shY2) H1299 cells. c Genome-browser view of the H3K9ac and ZZZ3 ChIP-seq peaks on the indicated ribosomal protein genes in cells as in (b). d qPCR analysis of H3K9ac ChIP of the indicated ribosomal protein genes in cells as in (b). e qPCR analysis of ZZZ3 ChIP of the indicated ribosomal protein genes in cells as in (b). f Average ZZZ3 occupancy on the promoter (5 kb±TSS) of the ZZZ3-bound genes or non-ZZZ3-bound genes (others) in control (shNT) and YEATS2 KD (shY2) H1299 cells. In d and e, error bars indicate S.E.M. of at least three biological replicates. *p < 0.05; **p < 0.01 (Student’s t-test) with longer chains. Indeed, recently we found that AF9, YEATS2, active transcription. However, the abundances of these newly and yeast Taf14 proteins are capable of binding to a repertoire of identified histone modifications are at levels of orders of mag- histone acylations, with slightly higher affinities to nitude lower than that of acetylation in cells33, 34, raising the – crotonylation28 30. Crotonylation and other types of acylations, question how cells discriminate these chemically closely related such as propionylation, butyrylation, and β-hydroxybutyrylation, modifications. In the current study, we failed to detect H3K27cr – have been detected on histones in a variety of species31 33. Similar ChIP-seq signals, likely due to the low abundance of this mark in to acetylation, these modifications on histones are associated with H1299 cells under normal growth conditions. Interestingly,

NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4

a N.S. N.S. N.S. N.S. N.S. 12 4 3 1.5 15 ** ** ** * * * ** ** * * shNT+Vector shY2+Vector 3 shY2+YEATS2 8 2 1.0 10 shY2+Y262A 2 shY2+W282A 4 1 0.5 5 1 H3K9ac ChIP (Relative to H3) (Relative 0 0 0 0.0 0 RPL8 RPL27 RPL29 RPL35 RPL38 b c N.S. N.S. N.S. N.S. N.S. 80 shNT+Vector shY2+Vector

1.2 ** ** ** ** ** ******** ) 4 60 shY2+YEATS2

** 10 0.9 × shY2+Y262A N.S. shY2+W282A 40 ** 0.6 * 0.3 20 Cell number ( Cell number Relative mRNA Level Relative 0.0 0 RPL8 RPL27 RPL29 RPL35 RPL38 01234 Days d f shNT+VectorshY2+Vector shY2+YEATS2 shY2+Y262A shY2+W282A 1.5 shNT+Vector shY2+Vector shY2+YEATS2 )

3 shY2+Y262A 1.0 shY2+W282A N.S.

** e 0.5 *

shNT+VectorshY2+Vector shY2+YEATS2 shY2+Y262A shY2+W282A (cm size Tumor

0.0 0123456 *** *** 350 160 *** *** Weeks post injection N.S. N.S. 300 m) μ 120 g 200 80 ATAC

100 40 YEATS2 GCN5 SGF29 H3K9 Colony diameter ( Colony AC H3K4 Colony number per field number Colony 0 0 Me AC H3K27

shNT+VectorshY2+VectorshY2+Y262A shNT+VectorshY2+VectorshY2+Y262A shY2+YEATS2shY2+W282A shY2+YEATS2shY2+W282A Ribosomal protein genes

Fig. 6 The YEATS domain of YEATS2 is required for ATAC-dependent ribosomal protein gene expression and tumor cell survival. a qPCR analysis of H3K9ac ChIP in the promoters of the indicated ribosomal protein genes in control (shNT) and YEATS2 KD (shY2) H1299 cells ectopically expressing shRNA-resistant WT YEATS2 or the indicated mutants. b qRT-PCR analysis of the expression of ribosomal protein genes in cells as in (a). In a and b, error bars indicate S.E.M. of at least three biological replicates. N.S. not significant; *p < 0.05; **p < 0.01 (Student’s t-test). c Cell proliferation assay of cells as in (a). Cells (mean ± S.E.M., n = 3) were counted for 4 days after seeding. Error bars represent the S.E.M. N.S.; *p < 0.05; **p < 0.01 (Student’s t-test). d Clonogenic assay of cells as in (a). Colonies were stained and photographed 7 days after seeding. e Anchorage-independent growth assay of cells as in (a). Cells (mean ± S.E.M., n = 4–6) were stained and photographed 3 weeks after seeding. Colony numbers (bottom left) and diameters (bottom right) were measured using ImageJ software. Error bars represent the S.E.M. Scale bar, 200 µm. N.S.; ***p < 0.001 (Student’s t-test). f Volumes of tumors (mean ± S.E.M., n = 10) of the H1299 cells as in (a) subcutaneously transplanted into immunodeficient nude mice. Tumors were monitored for 5 weeks after transplantation. N.S.; *p < 0.05; **p < 0.01 (Student’s t-test). g Working model: the YEATS2 subunit of the ATAC complex recognizes H3K27ac through its YEATS domain and stabilizes the ATAC complex at target promoter regions to maintain local histone acetylation and gene expression, which are essential for cell growth and survival. Note that additional reader modules, such as the SGF29 double Tudor domains that bind to H3K4me3, also contribute to chromatin association of the ATAC complex

10 NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4 ARTICLE several recent studies suggest that these alternative acylations YEATS2 (aa 184–449 or aa201–332) were cloned into the pGEX-6P1 vectors likely play a prominent role controlling gene expression at spe- (Novagen). Point mutations were generated using a site-directed mutagenesis kit – fi cific developmental stages or in responses to certain stresses33 35. (Stratagene). Histone peptides bearing different modi cations were synthesized at the W.M. Keck Facility at Yale University or CPC Scientific Inc. Anti-histone Nevertheless, given the preponderant abundance, histone acet- antibodies including anti-H3 (Ab1791, WB 1:20000), anti-H3K9ac (Ab32129, WB ylation likely plays a dominant role in epigenetic regulation of 1:1000), anti-H3K14ac (Ab52946, WB 1:1000), anti-H3K27ac (Ab4729, WB gene expression under most circumstances. 1:1000), anti-H4 (Ab731, WB 1:5000), and anti-HDAC1(ab19845, WB 1:1000) Acetylation on histone H3 K9 and K14 are known as marks for antibodies were obtained from Abcam; anti-H3K9ac (61251) and anti-H4K16ac 36 (39167 WB 1:1000) from Active Motif; anti-H4 tetra-acetyl antibody (06-598, active transcription . In mammals, acetylation on these two WB 1:5000) from Millipore; anti-GCN5 (sc-20698, WB 1:2000), anti-PCAF residues is predominantly deposited by GCN5/PCAF present in (sc-13124, WB 1:200), anti-ADA3 (sc-98821, WB 1:1000), anti-ATAC2 (sc-398475, SAGA or ATAC, two HAT complexes with non-overlapping WB 1:1000), and anti-GST (sc-459, WB 1:1000) antibodies from Santa Cruz; functions14, 15, 20. SAGA is principally found at gene promoters in anti-actin (A1978, WB 1:5000), anti-ZZZ3 (SAB4501106, WB 1:1000), and general whereas ATAC occupies both promoter and enhancer anti-tubulin (T8328, WB 1:5000) antibodies from Sigma; and anti-YEATS2 18 (24717-1-AP, WB 1:1000) antibody from ProteinTech. SGF29 antibody regions in GM12878 lymphoblast cells and HeLa cells . How- (WB 1:1000) and ChIP-seq grade ZZZ3 antibody were made in Dr. Laszlo ever, to our surprise, we did not observe any significant enhancer Tora’s laboratory18. pLKO shRNA constructs were purchased from Sigma. occupancy of the ATAC complex in the NSCLC H1299 cells, The shRNA sequences were: YEATS2#1: GCACAGAAACTGACTTCTTTA; indicating that the enhancer occupancy of the ATAC complex in YEATS2#2: TCAAAGAACTTGGTCATAAAT. different cell types is likely context-dependent. Moreover, other – reader modules within the complexes may also contribute the Protein production. The YEATS2 domain encompassing residues 201 332 of fi human YEATS2 was cloned into a pSUMOH10 vector (an in house modified binding speci city. In line with this speculation, SGF29, a shared vector based on pET28b) containing an N-terminal 10×His-SUMO tag. The subunit of the SAGA and ATAC complexes, has been shown to recombinant YEATS2201–332 was overexpressed in Escherichia coli BL21 (DE3). recognize H3K4me3 through its double Tudor domains37, 38, and After overnight induction by 0.4 mM isopropyl β-D-thiogalactoside at 16 °C in TB – the BRDs of GCN5 and PCAF bind to acetylated histones39 41. medium, cells were collected and suspended in buffer: 20 mM Tris, pH 7.5, 0.5 M sodium citrate, 5% glycerol, 1 mM phenylmethylsulfonyl fluoride, and 20 mM Therefore, YEATS2 likely cooperates with SGF29, GCN5/PCAF, imidazole. After cell lysis and centrifugation, the recombinant protein was purified and possibly some other yet unknown reader(s) to form a “reader to homogeneity over HisTrap, and the 10xHis-SUMO tag was cleaved by ULP1 module” within the ATAC complex facilitating chromatin overnight at 4 °C then removed by reloaded onto the HisTrap column. The free fi recruitment of the complex. As both SGF29 and GCN5/PCAF are YEATS2201–332 protein was nally polished by size-exclusion chromatography on a fi Superdex G75 column (GE Healthcare) in elution buffer: 20 mM Tris, pH 7.5, shared subunits of SAGA and ATAC, the ATAC-speci c YEATS2 0.5 M sodium citrate, 5% glycerol, and 2 mM β-mercaptoethanol. All fi and promoter H3K27ac may account for, at least in part, the YEATS2201–332 mutants were puri ed in essentially the same procedures as the differential distributions of SAGA and ATAC in promoters. wild-type protein. All mutant proteins were expressed and purified essentially the Furthermore, given that H3K27ac, but not H3K4me3, is enriched same as WT YEATS2 YEATS. For NMR titrations, the YEATS domain (aa 201–350) of YEATS2 was expressed in BL21(DE3) RIL cells as a GST-fusion in active enhancers, we speculate that the enhancer occupancy of 15 protein in minimal media supplemented with NH4Cl (Sigma). Cells were pelleted the ATAC complex in other cells likely depends on the func- via centrifugation, flash-frozen in liquid nitrogen, and lysed by sonication. Cell tionality of the YEATS2 YEATS domain rather than the SGF29 lysate was centrifuged, and the supernatant was incubated with glutathione Tudor domains. Sepharose 4B beads (GE Healthcare). The GST tag was cleaved with Prescission 15 The SAGA complex is known to play an essential role for both protease. The N-labeled YEATS2 YEATS domain was concentrated in 1× PBS (pH 6.8) buffer supplemented with 100 mM KCl before NMR experiments. normal and neoplastic development20. Components of the SAGA complex directly interact with the Myc oncoprotein and a ple- Crystallization and structure determination – – thora of transcription factors regulating gene expression involving . For YEATS2201 332 H3K27ac 42–44 complex, the sample was prepared by direct mixing protein with a H324–31K27ac in diverse processes . In contrast, little is known about the (ATKAARKacSAPA) in a 1:10 molar ratio. The crystals were generated by sitting pathways that ATAC complex is involved. In Drosophila, the drop vapor diffusion method. Briefly, protein droplets containing 1 μlof – μ ATAC complex serves as a transcriptional cofactor for c-Jun- YEATS2201–332 H3K27ac (6.5 mg/ml) were mixed with 1 l of reservoir solution regulating JNK target genes19, and in mammals, ATAC activates (0.2 M lithium sulfate, 2.0 M ammonium sulfate, 0.1 M 3-(cyclohexylamino)-1- 15, 45 propanesulfonic acid (CAPS), pH 10.5) and incubated in a closed 48-well plate at gene expression during stress responses . In the current study, 18 °C for 3 days. The crystals were then collected and briefly soaked in a cryo- we find that the ATAC complex also transcriptionally regulates a protectant drop composed of the reservoir solution supplemented with 30% gly- large number of essential genes including the ribosomal protein- cerol and then flash frozen in liquid nitrogen for data collection. The diffraction encoding genes. The YEATS2 gene is frequently amplified in data set was collected at the beamline BL17U of the Shanghai Synchrotron Radiation Facility at 0.9791 Å. All diffraction images were indexed, integrated, and NSCLC, especially the squamous sub-type. Knockdown of merged using HKL200048. The structure was determined by molecular replacement YEATS2 dampens the expression of 49 out of the total 79 ribo- using MOLREP49 with the AF9 complex structure (PDB ID: 4TMP) as the search somal protein genes and suppresses the growth and survival of a model. Structural refinement was carried out using PHENIX50, and iterative model 51 panel of lung cancer cells, suggesting a growth dependency of building was performed with COOT . Detailed data collection and refinement statistics are summarized in Table 1. Structural figures were created using the NSCLC on YEATS2, and possibly the ATAC complex. Interest- PYMOL (http://www.pymol.org/) or Chimera (http://www.cgl.ucsf.edu/chimera) ingly, Myc is known to regulate the expression of genes encoding programs. ribosomal proteins and several other components of the protein 46, 47 synthetic machinery . It is of interest to determine in future Isothermal titration calorimetry. All calorimetric experiments of the wild type or studies whether the ATAC complex also directly interacts with mutant YEATS domain proteins were conducted at 15 °C using a MicroCal iTC200 the Myc oncoprotein, or whether ATAC cooperates with the instrument (GE Healthcare). The YEATS2201–332 samples were dialyzed in the β SAGA complex in transcriptionally regulating the protein syn- following buffer: 20 mM Tris 7.5, 0.5 M sodium citrate, 5% glycerol, and 2 mM - fi mercaptoethanol. Protein concentration was determined by absorbance spectro- thetic machinery. Taken together, the identi cations of YEATS2 scopy at 280 nm. Peptides (>95% purity) were quantified by weighing on a large as a histone acetylation reader and a candidate oncogene ampli- scale and then aliquoted and freeze-dried for individual use. Acquired calorimetric fied in NSCLC suggest that the YEATS domain may provide an titration curves were analyzed using Origin 7.0 (OriginLab) using the “One Set of ” fi attractive therapeutic target for treatment. Binding Sites tting model. Detailed peptide sequence information is summarized below: H315–39K27ac: APRKQLATKAARK(ac)SAPATGGVKKPH, H31–34K27un: ARTKQTARKSTGGKAPRKQLATKAARKSAPATGG, H31–15K9ac: ARTKQ- Methods TARK(ac)STGGKA. Materials. Human YEATS2 cDNA (NCBI Gene ID 55789) was cloned into pENTR3C, and subsequently cloned into pCDH destination vectors using Gateway NMR titrations. NMR experiments were carried out on a Varian INOVA techniques (Invitrogen). The cDNAs encoding the YEATS domains of human 600 MHz at 298 K. 1H, 15N heteronuclear single-quantum coherence (HSQC)

NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4 spectra were collected on 0.1 mM uniformly 15N-labeled YEATS domain of stained with 0.005% crystal violet blue and photographed. Colony numbers and μ YEATS2 (aa 201–350 of YEATS2) in 1× PBS (pH 6.8) supplemented with 100 mM colony diameters were measured using ImageJ software with size cutoff of 15 m. Results were quantitated from six views per sample of at least three independent KCl and ~8% D2O in the presence of increasing concentration of H3 peptides replicates. (H3K27ac21–31 or H3K9ac1–12).

Peptide pulldown assay and GST pulldown assay. An aliquot of 1 µg of bioti- Flow cytometry cell cycle analysis. Cells were harvested and single cell suspen- nylated histone peptides with different modifications were incubated with 1–2 µgof sion was prepared at 2 × 106 in 1 ml ice-cold PBS buffer. The cell suspension was GST-fused proteins in binding buffer (50 mM Tris-HCl 7.5, 250 mM NaCl, 0.1% added dropwise to 9 ml 70% ethanol for fixing. The samples were kept at least 2 h NP-40, 1 mM phenylmethyl sulphonyl fluoride (PMSF)) at 4 °C overnight. Strep- at 4 °C then washed in cold PBS twice. Then the cells were treated with 100 μg/ml tavidin beads (Amersham) were added to the mixture, and the mixture was RNase A in PBS for 20 mins at room temperature, followed by adding propidium incubated for 1 h with rotation. The beads were then washed three times and iodide (PI; 50 μg/ml) for staining. The cell cycle profiling was analyzed by flow analyzed using SDS-PAGE and western blotting. For GST pulldown, 2 µg protein cytometry using 488 nm excitation. were incubated with 10 µg of calf thymus total histones (Worthington) in binding buffer (50 mM Tris-HCl 7.5, 1 M NaCl, 1% NP-40, 0.5 mM EDTA, 1 mM PMSF FITC Annexin V apoptosis assay. Phosphatidylserine (PS) translocation from the plus protease inhibitors (Roche)) at 4 °C overnight, followed by an additional 1 h inner to the outer leaflet of plasma membrane is one of the earliest apoptotic Glutathione Sepharose beads (Amersham) incubation. The beads were then washed features. The binding of Annexin V to cell surface PS was detected with a com- five times and analyzed using SDS-PAGE and western blotting. mercially available FITC Annexin V Apoptosis Detection Kit I (BD Pharmingen 556547). Briefly, 1 × 105 cells were pelleted, resuspended in 100 μl of Hepes- Protein-chromatin immunoprecipitation. Protein-ChIP assays for detection of buffered saline, and FITC-labeled annexin V and PI were added. The cells were YEATS2-histone interactions in cells were performed as described below52. Briefly, incubated 15 min at room temperature, then the samples were transferred to ice cells were crosslinked with 1% formaldehyde for 10 min and stopped with 125 mM and the sample volume brought to 0.5 ml. Analysis was done by flow cytometry glycine. The isolated nuclei were resuspended in nuclei lysis buffer and sonicated. within 1 h. The results were analyzed with FlowJo software. Annexin V positive The nuclei lysate was diluted in cell lysis buffer (20 mM Tris-HCl 8.0, 150 mM cells were determined as described in the Kit by setting quadrants to separate viable NaCl, 1% Triton X-100, 1 mM EDTA, 0.01% SDS, 1 mM PMSF plus protease cells from PI permeant cells, and non-apoptotic cells from those staining highly for inhibitors (Roche)). Anti-FLAG M2-conjugated agarose beads (Sigma) were the FITC-labeled Annexin V probe. Percent apoptosis was determined from the incubated with the lysates overnight at 4 °C. The beads were then washed with low cells staining greater than the control population threshold. salt (20 mM Tris pH 8.0, 150 mM NaCl, 2 mM EDTA, 1% Triton X-100, 0.1% SDS), high salt (20 mM Tris pH 8.0, 500 mM NaCl, 2 mM EDTA, 1% Triton X-100, Transwell cell migration assay. Cell migration was assayed using Transwell 0.1% SDS), and LiCl buffer (20 mM Tris pH 8.0, 250 mM LiCl, 1 mM EDTA, 1% μ NP-40, 1% sodium deoxycholate), and the bound proteins were eluted in SDS chambers (6.5 mm; Corning, Corning, NY, USA) with 8 m pore membranes. The lower chamber was filled with 500 μl of 10% FBS RPMI 1640 medium. A total of buffer and analyzed by western blotting. All uncropped blots are provided in 5 μ Supplementary Fig. 7. 1×10 cells were suspended with 500 l FBS-free RPMI 1640 medium and placed into the upper chamber. After 14 h, cells were fixed using 5% glutaraldehyde and stained using 0.5% crystal violet. Cells in the upper chamber were carefully Cell culture and RNA interference. All cell lines were tested for mycoplasma removed, and cells that migrated through the membrane were assessed by pho- contamination and validated by STR DNA fingerprinting performed by the tography. For quantification, crystal violet was extracted by methanol and the MDACC CCSG-funded Characterized Cell Line Core (NCI #CA016672). Human absorbance at 540 nm was measured. HEK 293 T, fibroblasts WI-38 and IM-R90 (ATCC), and human ovarian cancer cell lines CaoV3 and HeyA8 (gifts from Dr. Xiongbin Lu) were maintained in DMEM (Cellgro) supplemented with 10% fetal bovine serum (Sigma). Human lung ChIP and ChIP-seq analysis. ChIP analysis was performed essentially as described 53 fl cancer cell lines H1299, A549, H1355, Ludlu-1, and H520 (gifts from Dr. J. Hey- below . Brie y, cells were crosslinked with 1% formaldehyde for 10 min and mach) were cultured in RPMI 1640 (Cellgro) supplemented with 10% fetal bovine stopped with 125 mM glycine. The isolated nuclei were resuspended in nuclei lysis serum. Retroviral or lentiviral transduction was performed as described below53. buffer and sonicated using a Bioruptor Sonicator (Diagenode). The samples were – μ Briefly, 293 T cells were co-transfected with pMD2.G, pPAX2 (Addgene), and immunoprecipitated with 2 4 g of the appropriate antibodies overnight at 4 °C. pLKO shRNA or pCDH cDNA constructs. For infections, cells were incubated with Protein A/G beads were added and incubated for 1 h, and the immunoprecipitates viral supernatants in the presence of 8 µg/ml polybrene. After 48 h, the infected were washed twice each with low salt, high salt, and LiCl buffers. Eluted DNA was fi fi cells were selected with puromycin (2 µg/ml) for pLKO clones or blasticidin (10 µg/ reverse-crosslinked, puri ed using PCR puri cation kit (Qiagene), and analyzed by ml) for pCDH clones for 3–4 days before experiments. quantitative real-time PCR on the ABI 7500-FAST System using the Power SYBR Green PCR Master Mix (Applied Biosystems). Statistic differences were calculated using a two-way unpaired Student’s t-test. The primers used for qPCR are listed in Real-time PCR and RNA-seq analysis. Total RNA was extracted using an RNeasy the Supplementary Table 2. plus kit (Qiagen) and reverse-transcribed using an iScrip reverse transcription kit For ChIP-seq, ChIP experiments were carried out essentially the same as (Bio-Rad). Quantitative real-time PCR (qPCR) analyses were performed as described above. Samples were sequenced using the Illumina Solexa Hiseq 2500. described previously using Power SYBR Green PCR Master Mix and the ABI 7500- 53 The raw reads were mapped to human reference genome NCBI 37 (hg19) by Solexa FAST Sequence Detection System (Applied Biosystems) . Gene expressions were data processing pipeline, allowing up to 2 mismatches. The genome ChIP-seq calculated following normalization to GAPDH levels using the comparative Ct profiles were generated using MACS 1.3.657 with only unique mapped reads. (cycle threshold) method. Statistic differences were calculated using a two-way Clonal reads were automatically removed by MACS. The ChIP-seq profiles were ’ unpaired Student s t-test. The primer sequences for qPCR are listed in Supple- normalized to 10,000,000 total tag numbers, and peaks were called at p ≤ 1e−8. The mental Table 2. ChIP-seq heatmap was drawn by the seqplots R package (http://github.com/ RNA-seq samples were sequenced using the Illumina Hiseq 2500, and raw reads przemol/seqplots). were mapped to the human reference genome (hg19) and transcriptome using the RNA-Seq unified mapper. Read counts for each transcript were calculated using HTseq v0.6.1 using default parameters54. Differential gene expression analyses were Tumor xenograft. All animal studies were in compliance with ethical regulations performed using the “exactTest” function in edgeR v3.055. analysis at the University of Texas MD Anderson Cancer Center. Female athymic nude was performed using the DAVID Bioinformatics Resource 6.756. The gene mice (age 6–8 weeks) were obtained from University of Texas MD Anderson expression heatmap was generated using pheatmap package in CRAN (https://cran. Cancer Center and animals were housed under pathogen-free conditions. Tumor r-project.org/package=pheatmap). The volcano plot was drawn by using ggplot2 xenograft assay was performed as described below53. Briefly, three million YEATS2 package (https://cran.r-project.org/package=ggplot2) in R computing knockdown H1299 cells stably expressing control pCDH vector, wild-type environment. YEATS2, Y262A, or W282A mutants were suspended in 100 μl of serum-free RPMI 1640 and injected subcutaneously into the mice. The growth of tumors was monitored twice a week until the largest one reached the limit of tumor burden. Cell proliferation and colony formation assays . Cell proliferations were deter- Tumor sizes were measured using a caliper and tumor volume was calculated mined by counting live cells using hemocytometer cell counter or by CellTiter-Glo according to the following equation: tumor volume (mm3) = (length (mm) × luminescent cell viability assay kit (Progema-G7572). For colony formation assays, width2 (mm2)) × 0.5. Representative data were obtained from all the mice per H1299 cells were seeded in 6-well tissue culture plates (400 cells/well) and grown – fi experimental group. Statistical analyses were performed with one-way repeated- for 10 14 days. Colonies were xed with glutaraldehyde (6.0% v/v), stained with measures analysis of variance. crystal violet (0.5% w/v) and photographed. Soft agar assays were performed as described below53. Briefly, cells (1 × 104) were suspended in 1 ml top agar medium (culture medium supplied with 0.35% Statistical analyses. Experimental data are presented as means ± standard agar). The cell suspensions were then overlaid onto 1.5 ml bottom agar medium deviation of the mean unless stated otherwise. Statistical significance was calculated (culture medium supplied with 0.6% agar) in six-well tissue culture plates in unless stated otherwise by two-tailed unpaired t-test on two experimental condi- triplicate. Fresh medium was added to plates every 3 days. On day 21, cells were tions with p < 0.05 considered statistically significant. Statistical significance levels

12 NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4 ARTICLE are denoted as follows: *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. No sta- 24. Luo, Z. J., Lin, C. Q. & Shilatifard, A. The super elongation complex (SEC) tistical methods were used to predetermine sample size. Super exact test was family in transcriptional control. Nat. Rev. Mol. Cell. Biol. 13, 543–547 (2012). performed to test the significance of the Venn diagram by the R package Exact 25. Nguyen, A. T. & Zhang, Y. The diverse functions of Dot1 and H3K79 (https://cran.r-project.org/package=Exact). . Genes Dev. 25, 1345–1358 (2011). 26. Cai, Y. et al. Identification of new subunits of the multiprotein mammalian Data availability. Structure data are deposited in the with the TRRAP/TIP60-containing histone acetyltransferase complex. J. Biol. Chem. – accession code 5XNV. The ChIP-seq and RNA-seq data are deposited in the GEO 278, 42733 42736 (2003). database with the accession number GSE90781. 27. Cai, Y. et al. The mammalian YL1 protein is a shared subunit of the TRRAP/ TIP60 histone acetyltransferase and SRCAP complexes. J. Biol. Chem. 280, 13665–13670 (2005). Received: 23 December 2016 Accepted: 24 August 2017 28. Andrews, F. H. et al. The Taf14 YEATS domain is a reader of histone crotonylation. Nat. Chem. Biol. 12, 396–398 (2016). 29. Li, Y. et al. Molecular coupling of histone crotonylation and active transcription by AF9 YEATS domain. Mol. Cell 62, 181–193 (2016). 30. Zhao, D. et al. YEATS2 is a selective histone crotonylation reader. Cell Res. 26, 629–632 (2016). References 31. Tan, M. et al. Identification of 67 histone marks and histone lysine 1. Strahl, B. D. & Allis, C. D. The language of covalent histone modifications. crotonylation as a new type of histone modification. Cell 146, 1016–1028 Nature 403,41–45 (2000). (2011). 2. Kouzarides, T. Chromatin modifications and their function. Cell 128, 693–705 32. Dai, L. et al. Lysine 2-hydroxyisobutyrylation is a widely distributed active (2007). histone mark. Nat. Chem. Biol. 10, 365–370 (2014). 3. Brown, C. E., Lechner, T., Howe, L. & Workman, J. L. The many HATs of 33. Xie, Z. et al. Metabolic regulation of gene expression by histone lysine beta- transcription coactivators. Trends Biochem. Sci. 25,15–19 (2000). hydroxybutyrylation. Mol. Cell 62, 194–206 (2016). 4. Jenuwein, T. & Allis, C. D. Translating the . Science 293, 34. Sabari, B. R. et al. Intracellular crotonyl-CoA stimulates transcription through 1074–1080 (2001). p300-catalyzed histone crotonylation. Mol. Cell 58, 203–215 (2015). 5. Dhalluin, C. et al. Structure and ligand of a histone acetyltransferase 35. Dai, L. Z. et al. Lysine 2-hydroxyisobutyrylation is a widely distributed active bromodomain. Nature 399, 491–496 (1999). histone mark. Nat. Chem. Biol. 10, 365–U73 (2014). 6. Zeng, L. et al. Mechanism and regulation of acetylated histone binding by the 36. Li, B., Carey, M. & Workman, J. L. The role of chromatin during transcription. tandem PHD finger of DPF3b. Nature 466, 258–262 (2010). Cell 128, 707–719 (2007). 7. Ali, M. et al. Tandem PHD Fingers of MORF/MOZ acetyltransferases display 37. Bian, C. B. et al. Sgf29 binds histone H3K4me2/3 and is required for SAGA selectivity for acetylated histone H3 and are required for the association with complex recruitment and histone H3 acetylation. EMBO J. 30, 2829–2842 chromatin. J. Mol. Biol. 424, 328–338 (2012). (2011). 8. Qiu, Y. et al. Combinatorial readout of unmodified H3R2 and acetylated H3K14 38. Vermeulen, M. et al. Quantitative interaction proteomics and genome-wide by the tandem PHD finger of MOZ reveals a regulatory mechanism for HOXA9 profiling of Epigenetic histone marks and their readers. Cell 142, 967–980 transcription. Genes Dev. 26, 1376–1391 (2012). (2010). 9. Li, Y. et al. AF9 YEATS domain links histone acetylation to DOT1L-mediated 39. Hudson, B. P., Martinez-Yamout, M. A., Dyson, H. J. & Wright, P. E. Solution H3K79 methylation. Cell 159, 558–571 (2014). structure and acetyl-lysine binding activity of the GCN5 bromodomain. J. Mol. 10. Le Masson, I. et al. Yaf9, a novel NuA4 histone acetyltransferase subunit, is Biol. 304, 355–370 (2000). required for the cellular response to spindle stress in yeast. Mol. Cell Biol. 23, 40. Zeng, L., Zhang, Q., Gerona-Navarro, G., Moshkina, N. & Zhou, M. M. 6086–6102 (2003). Structural basis of site-specific histone recognition by the bromodomains of 11. Schulze, J. M., Wang, A. Y. & Kobor, M. S. YEATS domain proteins: a diverse human coactivators PCAF and CBP/p300. Structure 16, 643–652 (2008). family with many links to chromatin modification and transcription. Biochem. 41. Filippakopoulos, P. et al. Histone recognition and large-scale structural analysis Cell Biol. 87,65–75 (2009). of the human bromodomain family. Cell 149, 214–231 (2012). 12. Wang, Y. L., Faiola, F., Xu, M., Pan, S. & Martinez, E. Human ATAC Is a 42. Zhang, N. et al. MYC interacts with the human STAGA coactivator complex via GCN5/PCAF-containing acetylase complex with a novel NC2-like histone fold multivalent contacts with the GCN5 and TRRAP subunits. Biochim. Biophys. module that interacts with the TATA-binding protein. J. Biol. Chem. 283, Acta 1839, 395–405 (2014). 33808–33815 (2008). 43. Liu, X., Tesfai, J., Evrard, Y. A., Dent, S. Y. & Martinez, E. c-Myc 13. Guelman, S. et al. Host cell factor and an uncharacterized SANT domain transformation domain recruits the human STAGA complex and requires protein are stable components of ATAC, a novel dAda2A/dGcn5-containing TRRAP and GCN5 acetylase activity for transcription activation. J. Biol. Chem. histone acetyltransferase complex in Drosophila. Mol. Cell Biol. 26, 871–882 278, 20405–20412 (2003). (2006). 44. Hirsch, C. L. et al. Myc and SAGA rewire an alternative splicing network during 14. Spedale, G., Timmers, H. T. & Pijnappel, W. W. ATAC-king the complexity of early somatic cell reprogramming. Genes Dev. 29, 803–816 (2015). SAGA during evolution. Genes Dev. 26, 527–541 (2012). 45. Sela, D. et al. Endoplasmic reticulum stress-responsive transcription factor 15. Nagy, Z. et al. The metazoan ATAC and SAGA coactivator HAT complexes ATF6alpha directs recruitment of the Mediator of RNA polymerase II regulate different sets of inducible target genes. Cell. Mol. Life Sci. 67, 611–628 transcription and multiple histone acetyltransferase complexes. J. Biol. Chem. (2010). 287, 23035–23045 (2012). 16. Suganuma, T. et al. ATAC is a double histone acetyltransferase complex that 46. Barna, M. et al. Suppression of Myc oncogenic activity by ribosomal protein stimulates sliding. Nat. Struct. Mol. Biol. 15, 364–372 (2008). haploinsufficiency. Nature 456, 971–U79 (2008). 17. Riss, A. et al. Subunits of ADA-two-A-containing (ATAC) or Spt-Ada-Gcn5- 47. van Riggelen, J., Yetil, A. & Felsher, D. W. MYC as a regulator of ribosome acetyltrasferase (SAGA) coactivator complexes enhance the acetyltransferase biogenesis and protein synthesis. Nat. Rev. Cancer 10, 301–309 (2010). activity of GCN5. J. Biol. Chem. 290, 28997–29009 (2015). 48. Otwinowski, Z. & Minor, W. Processing of X-ray diffraction data collected in 18. Krebs, A. R., Karmodiya, K., Lindahl-Allen, M., Struhl, K. & Tora, L. SAGA and oscillation mode. Methods Enzymol. 276, 19 (1997). ATAC histone acetyl transferase complexes regulate distinct sets of genes and 49. Vagin, A. & Teplyakov, A. Molecular replacement with MOLREP. Acta ATAC defines a class of p300-independent enhancers. Mol. Cell 44, 410–423 Crystallogr. D Biol. Crystallogr. 66,22–25 (2010). (2011). 50. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for 19. Suganuma, T. et al. The ATAC acetyltransferase complex coordinates MAP macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, kinases to regulate JNK target genes. Cell 142, 726–736 (2010). 213–221 (2010). 20. Wang, L. & Dent, S. Y. Functions of SAGA in development and disease. 51. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. Epigenomics 6, 329–339 (2014). Acta Crystallogr. D Biol. Crystallogr. 60, 2126–2132 (2004). 21. Lee, K. K. et al. Combinatorial depletion analysis to assemble the network 52. Shi, X. et al. ING2 PHD domain links histone H3 lysine 4 methylation to active architecture of the SAGA and ADA chromatin remodeling complexes. Mol. gene repression. Nature 442,96–99 (2006). Syst. Biol. 7, 503 (2011). 53. Wen, H. et al. ZMYND11 links histone H3.3K36me3 to transcription 22. Borowicz, S. et al. The soft agar colony formation assay. J. Vis. Exp. 92, e51998 elongation and tumour suppression. Nature 508, 263–268 (2014). (2014). 54. Anders, S., Pyl, P. T. & Huber, W. HTSeq-a Python framework to work with 23. Orpinell, M. et al. The ATAC acetyl transferase complex controls mitotic high-throughput sequencing data. Bioinformatics 31, 166–169 (2015). progression by targeting non-histone substrates. EMBO J. 29, 2381–2394 55. Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor (2010). package for differential expression analysis of digital gene expression data. Bioinformatics 26, 139–140 (2010).

NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-01173-4

56. Barr, A. J. et al. Large-scale structural analysis of the classical human protein Additional information tyrosine phosphatome. Cell 136, 352–363 (2009). Supplementary Information accompanies this paper at doi:10.1038/s41467-017-01173-4. 57. Zhang, Y. et al. Model-based analysis of ChIP-Seq (MACS). Genome Biol. 9, R137 (2008). Competing interests: W.M. is a M.D. Anderson Center for Cancer Epigenetics postdoctoral scholar. D.Z. is a postdoctoral fellow of Tsinghua-Peking Joint Center for Life Sciences. F.H.A. is an AHA postdoctoral fellow. W.L. is a fellow of the Jane Coffin Acknowledgements Childs Memorial Fund. X.S. is a Leukaemia & Lymphoma Society Career Development We thank J. Heymach, J. Minna, J. Kurie, M. Bedford, M.G. Lee, and X. Lu for sharing Program Scholar and a R. Lee Clark Fellow and Faculty Scholar of M.D.. Anderson reagents. We thank the M.D. Anderson Sequencing and Microarray Facility and the Cancer Center. X.S. is a Scientific Advisory Board member of EpiCypher. The remaining Science Park Next-Generation Sequencing Facility (CPRIT RP120348) for Solexa authors declare no competing financial interests. sequencing. We thank the staff at beamline BL17U of the Shanghai Synchrotron Radiation Facility and Dr. S. Fan at Tsinghua Center for Structural Biology for their Reprints and permission information is available online at http://npg.nature.com/ assistance in data collection and the China National Center for Protein Sciences Beijing reprintsandpermissions/ for providing facility support. We thank B. Dennehey for editing the manuscript. This work was supported in part by grants from NIH/NCI (1R01CA204020-01), Cancer Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in Prevention and Research Institute of Texas (RP160237 and RP140323), Welch Foun- published maps and institutional affiliations. dation (G1719), and Texas Tobacco Settlement to X.S., CPRIT (RP110471 and RP150292) and NIH (R01HG007538 and R01CA193466) to W.L., NIH (R01GM100907) to T.G.K., NIH (R01GM067718) to S.Y.R.D., National Natural Science Foundation of China (91519304), Major State Basic Research Development Program in China Open Access This article is licensed under a Creative Commons (2015CB910503), and Tsinghua University Initiative Scientific Research Program to H.L., Attribution 4.0 International License, which permits use, sharing, National Postdoctoral Program for Innovative Talents (BX201600088) to D.Z., and adaptation, distribution and reproduction in any medium or format, as long as you give European Research Council (ERC) (ERC-2013-340551, Birtoaction) to L.T. appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Author contributions indicated otherwise in a credit line to the material. If material is not included in the W.M., H.G., J.L. and D.Z. contributed equally to this work. X.S., H.L., W.L. and W.M. article’s Creative Commons license and your intended use is not permitted by statutory conceived the study. W.M. performed the biochemical and cellular studies; H.G. and D.Z. regulation or exceeds the permitted use, you will need to obtain permission directly from performed structural and calorimetric studies; J.L. and Y.X. performed bioinformatics the copyright holder. To view a copy of this license, visit http://creativecommons.org/ analysis; S.J. performed xenograft assays, F.H.A. performed NMR titration studies; M.G. licenses/by/4.0/. performed pathological analysis of tumors; X.W. provided technical assistance; L.T. provided ChIP-grade anti-ZZZ3 polyclonal antibody; X.S., H.L. and W.M. wrote the paper with comments from H.W., S.Y.R.D., L.T., J.L. and W.L. © The Author(s) 2017

14 NATURE COMMUNICATIONS | 8: 1088 | DOI: 10.1038/s41467-017-01173-4 | www.nature.com/naturecommunications